Scholarlink

Driver fatigue is a major cause of traffic accidents worldwide. This paper proposes a short-term fatigue monitoring approach with dual optimization of window size and features. A multi-modal data collection platform was built using a driving simulator to collect data from 30 participants, including PPG, EDA, facial videos, and vehicle motion. After processing, 34 multi-modal features were extracted from physiological, appearance, and behavioral dimensions. A genetic algorithm jointly optimized the time window and feature subset. Results show that a 10-s window with selected features (AVHR, NN50, SampleEn, PERCLOS, blink frequency) achieved an AUC of 0.97 and response time of 10.005 s, providing an optimal trade-off between timeliness and accuracy.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Research on Dual Optimization of Short-Term Fatigue Monitoring Window and Feature Based on Multimodal Coupling Analysis

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Liu Jia,
  • Zheng Xue-lian,
  • Ren Yuan-yuan,
  • Wang Qing-ju,
  • Yu Shang-ting,
  • Li Xian-sheng

摘要

Driver fatigue is a major cause of traffic accidents worldwide. This paper proposes a short-term fatigue monitoring approach with dual optimization of window size and features. A multi-modal data collection platform was built using a driving simulator to collect data from 30 participants, including PPG, EDA, facial videos, and vehicle motion. After processing, 34 multi-modal features were extracted from physiological, appearance, and behavioral dimensions. A genetic algorithm jointly optimized the time window and feature subset. Results show that a 10-s window with selected features (AVHR, NN50, SampleEn, PERCLOS, blink frequency) achieved an AUC of 0.97 and response time of 10.005 s, providing an optimal trade-off between timeliness and accuracy.